any Anvil user noticing builds breaking after any change to dagger code? (say adding a parameter in ...

# dagger

[ursus]

any Anvil user noticing builds breaking after any change to dagger code? (say adding a parameter in @Provides functions), fails with
...:compileWithJavac
Cannot find symbol FooModule_XYZFactory

cant pinpoint if its dagger or anvil causing this ..

[wasyl]

Definitely anvil, incremental compilation changes aren't picked up

[wasyl]

There's an open issue about multibindings changes not being picked up, too. Rerunning kotlin compilation task for that module with the failing class fixes the issue

[ursus]

clean build fixes it yea but its costing me so much time

[wasyl]

I'm saying it's Anvil because it started happening for us after we switched to anvil to generate dagger modules. And it's changes to anvil-only module that don't properly regenerate classes

[ursus]

how recent is that? because im anvil user since1.0 (or whenever they added the ability to generate dagger factories) and this for sure wasnt happening

[wasyl]

We migrated some time this year? Couple of months back at most. It might correlate with Kotlin 1.9

[ursus]

so you just clean build all the time?

[wasyl]

Rerun the tasks when it happens, yes 😢 although it's not happening constantly. It is annoying though, yeah.
I wanted to open a new issue for both anvil and JB but haven't gotten around to writing a proper repro

[ursus]

I presume this doesnt happen on hello world projects..

[wasyl]

Haven't tried 😅 for me it's definitely related to switching branches and pulling new changes from vcs

[ralf]

No need to file a new ticket, it’s covered in https://github.com/square/anvil/issues/693

This issue started to happen with Kotlin’s new incremental compiler and Kotlin 1.8.22 and later.

[ursus]

is there a more lightweight solution to workaround it other than a clean build?

that is killing me, since it's happening to me on any dagger change, not just switching branches etc

[wasyl]

You can rerun a single compilation task, ./gradlew failingModule:compileKotlin --rerun. If you have more modules it's not that heavy

[wasyl]

Ralf is there an issue I can track on YouTrack for this?

[ralf]

These 3 issues in particular:
• https://youtrack.jetbrains.com/issue/KT-57919/Store-all-annotations-in-Kotlin-metadata-on-JVM
• https://youtrack.jetbrains.com/issue/KT-51733/Provide-incremental-compilation-API-for-compiler-plugins
• https://youtrack.jetbrains.com/issue/KT-54850/Provide-mechanism-for-compiler-plugins-to-add-custom-information-into-binaries

[ursus]

when ksp dagger comes online, will it alleviate the issue? i.e. when ksp dagger, then one can remove the

anvil {
    generateDaggerFactories.set(true)
}

?

generateDaggerFactories is basically just a optimization to not run kapt everywhere, right?

[wasyl]

Since it uses the same ksp infrastructure as anvil (?), I feel it might have the same problem

[ursus]

idk but I feel like anvil generatedaggerblabla doesnt generate kotlin source as ksp does, but is one level deeper

[ralf]

No, Dagger KSP doesn’t have this issue, because they use a separate Kotlin compilation task.

[ralf]

It’s the opposite, Anvil generates Kotlin but Dagger KSP generates Java

[ursus]

btw how will that play together in the future? will generateDaggerFactories become deprecated and anvil will send people to use ksp dagger?

(and anvil just do anvil multimodule-meta-dagger stuff it's advertising?)

[ralf]

Not sure yet. Keep in mind that generateDaggerFactories is totally optional and only a build performance optimization

[ursus]

yea hence me asking if its obsolete once kapt is ritually murdered 😄

[Joel Wilcox]

It's possible -- we'll likely do some benchmarks to see if there are still any performance gains with generateDaggerFactories before making any decisions.